rageek

A place for Unix Thoughts and Ideas

Monthly Archives: October 2011

Solaris CAT (Crash Analysis Tool) is back

Oracle has reinstated the Solaris Crash Analysis Tool (Solaris CAT) and it is downloadable through metalink.

Here are the relevant patch numbers:

10364261 Solaris Crash Analysis Tool 5.2 Oracle Solaris on SPARC (32-bit)

10364269 Solaris Crash Analysis Tool 5.2 OpenSolaris (American English)

10364262 Solaris Crash Analysis Tool 5.2 Oracle Solaris on SPARC (64-bit)

10364266 Solaris Crash Analysis Tool 5.2 Oracle Solaris on x86 (32-bit)

10364267 Solaris Crash Analysis Tool 5.2 Oracle Solaris on x86-64 (64-bit)

Or you can search for “Solaris Crash Analysis Tool”

Here is a link to a useful writeup on Cuddletech for Solaris CAT:

http://www.cuddletech.com/blog/pivot/entry.php?id=966

Cleaning up zone package management locks

While trying to deploy OEM Ops Center, I was noticing that my deploys were hanging on my large global zone servers.

On investigation, I noticed that stale package locks in my local zones were preventing the package installation from completing.

I’m sure these locks usually serve a important purpose, but I’m also aware that the package management framework hasn’t been bullet proof with Solaris 10 zones.

The solution, was the kill the installer, clear the locks and re-run the installation.

Here is how you find the locks:

List the locks
# ls /zones/*/root/tmp/.ai.pkg.zone.lock-*

Remove the locks
# rm /zones/*/root/tmp/.ai.pkg.zone.lock-*

BTW, before removing the locks, ,ake sure there isn’t any package operations going on in any of the zones prior, otherwise strange things may happen.

Recovering from prtdiag hanging on CMT boxes.

Occasionally I will run into a situation on my RAC clusters (and systems with Sun Connect) where I will see and hundreds of prtdiag processes hanging in the process listing. This will cause the system load to just explode.

I have seen this occur on T2000’s and T5120’s.

You could run pkill to clean them up, but that won’t fix the problem.

The problem is that PICL is hanging and this is fixed by restarting it with:
svcadm restart picl

After restarting picl, all the prtdiag processes should complete.